This is an update of a Cochrane review first published in The Cochrane Library in Issue 4, 2006.
Otitis media with effusion (OME) or 'glue ear' is an accumulation of fluid in the middle ear, in the absence of acute inflammation or infection. It is the commonest cause of acquired hearing loss in childhood (Rovers 1999). It is common in children between the ages of one and three years and in seasons when there is a high incidence of the common cold, with a prevalence of OME between 10% and 30%. The cumulative incidence to the age of four years is 80% (Zielhuis 1990) and at age seven the prevalence is still 3% to 8% (Fiellau 1977; Fiellau 1983; Lous 1981; Teele 1989). The aetiology of 'glue ear' is uncertain, but low-grade infection, poor Eustachian tube function and adenoidal infection or hypertrophy have all been implicated (Bluestone 1988). Otitis media with effusion often resolves spontaneously within a few months (Fiellau 1979).
Otitis media with effusion (OME) may be associated with ear ache (otalgia), infection and/or hearing loss. Hearing loss may be significant (20 dB to 30 dB), particularly when the disorder is bilateral and has lasted for more than a month (Fiellau 1983). However, some children have nearly normal hearing despite the presence of fluid within the middle ear. The hearing loss and discomfort may have linguistic, developmental, behavioural and other social consequences, particularly if the effusion is bilateral and of long duration (Lous 1995). Some studies have shown that children with unilateral or bilateral otitis media with effusion have balance problems (Golz 1998; Grace 1990).
For its diagnosis, tympanometry in combination with otomicroscopy or pneumatic otoscopy is the recommended technique (Bluestone 1988). Tympanometry is performed with a handheld device inserted in the ear canal and measures the eardrum responses to the sound and different pressures. Diagnostic tympanometry results in a flat curve (relative gradient less than 0.1, type B) or in a curve with a middle ear pressure between -399 to -200 daPa (C2 curve) (Jerger 1970; Zielhuis 1989) or when mobility of the tympanic membrane is absent or reduced, or fluid and/or air bubbles are evident behind the eardrum (Browning 2010). Type A and C1 tympanograms are considered not to represent otitis media with effusion. The presence of a significant (10 dB) air-bone gap on pure-tone audiometry correlates well with the presence of fluid in the middle ear.
OME usually resolves spontaneously (20% will resolve in one month and 40% at three months (Fiellau 1979)) and many children will require no specific treatment. The most common medical treatment options include the use of decongestants (Griffin 2011), mucolytics, steroids, antihistamines (Griffin 2011) and antibiotics (van Zon 2012). Of these options, only steroids have been shown effective and only for short-term resolution (Simpson 2011). Surgical treatment is more widely used and better researched, with options including grommet insertion (Browning 2010), myringotomy tympanocentesis (surgical incision of the eardrum, with or without aspiration of fluid from the middle ear cavity) and adenoidectomy (van den Aardweg 2010). The optimal treatment strategy remains controversial, with wide international variability in clinical practice.
'Autoinflation' refers to the opening of, and forcing of air through, the Eustachian tube by the raising of intranasal pressure. This may be achieved by forced exhalation with closed mouth and nose, blowing up a balloon through each nostril, use of an anaesthetic mask or a Politzer device. The aim of these procedures is to introduce air into the middle ear, via the Eustachian tube, to equilibrate the pressures. This is most likely to help in 'Eustachian tube dysfunction' and has been proposed as a treatment for otitis media with effusion (Hunt-Williams 1968). This systematic review aims to assess the effectiveness of autoinflation devices in children and adults with otitis media with effusion.
To assess the effectiveness of autoinflation compared with no treatment in children and adults with otitis media with effusion.
Criteria for considering studies for this review
Types of studies
We included randomised controlled trials of an autoinflation device - designed to force air into the middle ear - in children or adults with clinically diagnosed serous otitis media.
Types of participants
Children and adults with unilateral or bilateral otitis media with effusion and a clinical diagnosis by primary care physicians or specialists using tympanometry (type B or C2), either alone or in combination with simple or pneumatic otoscopy or audiometry. The type of diagnosis used in each trial is presented in the table Characteristics of included studies.
Types of interventions
We included trials that studied any form of autoinflation (a technique to increase intranasal pressure repeatedly either through a nasal balloon or by other methods) compared to no autoinflation and which were unconfounded. By unconfounded, we mean studies where the two groups were treated equally, except for the provision of the intervention to one group. Other treatments (e.g. analgesia, decongestants or short courses of antibiotics) were permitted provided these were provided equally. Autoinflation interventions were classified into similar groups by an ENT surgeon, based on the description of the intervention whilst blinded to the outcomes of the study (see Appendix 1 for a brief description of the interventions).
Types of outcome measures
- Improvement in tympanogram (to type C1 or A tympanogram)
- Differences in hearing level on pure-tone audiogram (average 10 dB improvement over the frequencies 250 Hz to 2000 Hz)
- Improvement measured as a composite of change in tympanogram and/or audiometry (post hoc)
- Adverse effects of autoinflation
- Presence or absence of fluid in middle ear cavity
Time points of outcome assessment
- Up to and including one month from commencement of autoinflation
- More than one month from commencement of autoinflation
We were unable to analyse the following secondary outcomes due to a lack of trial data:
- Developmental test results
- Behavioural test results
- Language and speech development
- Quality of life
- Family function
Search methods for identification of studies
We conducted systematic searches for randomised controlled trials. There were no language, publication year or publication status restrictions. The date of the last search was 12 April 2013.
We searched the following databases from their inception for published, unpublished and ongoing trials: the Cochrane Ear, Nose and Throat Disorders Group Trials Register; the Cochrane Central Register of Controlled Trials (CENTRAL, The Cochrane Library 2013, Issue 3); PubMed; EMBASE; CINAHL; LILACS; KoreaMed; IndMed; PakMediNet; CAB Abstracts; Web of Science; ISRCTN; ClinicalTrials.gov; ICTRP; Google Scholar and Google.
We modelled subject strategies for databases on the search strategy designed for CENTRAL. Where appropriate, we combined subject strategies with adaptations of the highly sensitive search strategy designed by The Cochrane Collaboration for identifying randomised controlled trials and controlled clinical trials (as described in the Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0, Box 6.4.b. (Handbook 2011)). Search strategies for major databases including CENTRAL are provided in Appendix 2.
Searching other resources
We scanned the reference lists of identified publications for additional trials and contacted trial authors where necessary. In addition, we searched PubMed, TRIPdatabase, The Cochrane Library and Google to retrieve existing systematic reviews relevant to this systematic review, so that we could scan their reference lists for additional trials. We searched for conference abstracts using the Cochrane Ear, Nose and Throat Disorders Group Trials Register.
Data collection and analysis
Selection of studies
The Trials Search Co-ordinator of the Cochrane Ear, Nose and Throat Disorders Group assessed abstracts of all studies identified by the initial search and excluded clearly irrelevant studies. One review author assessed abstracts for relevance and we obtained full texts for these. Two review authors independently assessed all full-text articles. Differences in opinion about inclusion of studies were resolved by discussion with a third review author. The review authors were not blind to the journal of origin, the authors, the institutions or the magnitude of results.
Data extraction and management
Two review authors independently extracted study data using standardised forms. We entered data into RevMan 5 (RevMan 2012) and a second review author checked the data. Data extracted and presented included: author, publication year, journal, participants (numbers, duration of illness, demographics, characteristics of the disease, etc.), intervention (type of intervention and duration) and results (outcome measures, effect size, statistical significance, adverse effects). We contacted trial authors for missing or incomplete data.
Assessment of risk of bias in included studies
Two review authors independently assessed the risk of bias in all included trials. Differences were resolved by discussion with a third review author. We assessed risk of bias using the Cochrane 'Risk of bias' tool where types of bias are rated as low, high or unclear.
- Randomisation/selection bias: selection generation
- Randomisation/selection bias: allocation concealment
- Detection bias: blinding of outcome assessment
- Attrition bias: incomplete outcome data, losses to follow-up and whether analysis was intention-to-treat
- Blinding of outcome assessors
- Selective reporting
- Other bias: we also used this section to assess whether the groups were comparable at the start of the study for prognostic characteristics
We assessed the quality of the studies separately:
- that the groups were equally provided of care, except for the treatment; and
- adherence to study procedure.
We carried out analysis on the basis of intention-to-treat. Where possible we combined data to give a summary measure of effect.
When outcome data obtained were binary - recovered or improved glue ear or audiometry improvement - we used relative risks of improvement to compare interventions. The data reported in the trials mainly used 'ears' as their analysis unit despite using individuals as randomisation units. To correct for this and to give adequate weight to the studies we adjusted the outcomes using the 'design effect', which takes into consideration the association found in using both ears from the same individual (Deeks 2005).
We estimated the 'design effects' from the Blanshard data (full data set provided by author) at one month and three months for participants with either types B or C2 tympanograms or only type B tympanograms. Minor differences in correlation between intervention and control groups (smaller at three months) were found, but an average was used, as differences could be due to sampling variation. For participants with type B or C2 tympanograms the design effect at one month was 1.25, while at three months it was 1.51. For participants with only type B tympanograms the design effect was 1.3 for both one and three months follow-up.
We used the appropriate design effect (for less than one month or more than one month by tympanogram type at entry) to obtain adjusted risk ratios (RR) for all studies that provided data for the outcome. We calculated and then entered the log(RR) and the SE(log(RR)) into RevMan 5 (RevMan 2012). As the focus of the review is on recovery/improvement we emphasise the use of a positive outcome and refer to the RR obtained as the 'relative risk of improvement' or RRI; instead of using the term risk ratio which is normally associated with a negative outcome (e.g. death).
When data were continuous, we used mean differences to compare interventions. We used the GIV (generic inverse variance) method with a random-effects model to pool studies together and examined heterogeneity using the Chi
Subgroup analyses had been planned on the basis of the diagnosis of otitis media with effusion and the extent of the hearing loss measured on audiometry. However, insufficient studies were identified to allow these analyses. We also planned subgroups according to age (under three years, three to 12 years and over 12 years). However, we could not identify any papers of sufficient quality including children aged under three years and only one paper included children over 12 years (Lesinskas 2003).
Description of studies
Results of the search
From the 2013 update searches, we retrieved a total of 401 references. Following first-level screening (i.e. removal of duplicates and clearly irrelevant references) by the Cochrane ENT Group Trials Search Co-ordinator we were left with 127 references for further consideration. Following sifting by the authors two more duplicates were removed and 105 discarded following screening of titles and abstracts, leaving 20 reports for consideration. Two were studies which we have included in the updated review (De Nobili 2008; Ercan 2005). Sixteen further studies were excluded from the review, with reasons (Costantino 2010; Dalchow 2011; Ducla-Soares 2007; El Hachem 2012; Hidir 2011; Jumah 2010; Karahatay 2008; Kawase 2008; Laina 2006; Leach 2008; Pau 2009; Poe 2011; Prabhakar 2007; Shim 2010; Talmon 2010; Toros 2010). Two were identified as ongoing studies (Tel-Aviv Med Center 2006; Williamson 2011).
From the original searches (2006 publication of this review), 418 abstracts were obtained from the search, with 62 selected as possibly relevant. These were screened for relevance and we obtained 24 full-text articles. Of these full-text articles two review authors (JH and RP) filtered out 16 as duplicates or rejected them for not fulfilling the inclusion criteria. Two further reports were awaiting assessment in the original review and remain so (Niebuhr-Jorgensen 1996; Scadding 2002) (see Characteristics of studies awaiting classification). Limited available information has meant that these studies cannot yet be excluded from meeting the review criteria. There have been no published data following on from either of the abstracts for these studies. Contact with one of the authors has indicated that one of the studies may be published in the future (Scadding 2002), whilst the other author has not been contactable (Niebuhr-Jorgensen 1996). Six studies were included in the original (2006) version of the review.
It is hoped that the two ongoing studies will be able to contribute their results to this review in the future. The Tel-Aviv Med Center 2006 study commenced in October 2006. The authors confirmed in 2010 that their results would be released shortly; unfortunately this has been delayed and a new date has not been announced. The Williamson 2011 study started in September 2011 and it is hoped that results will be available during winter of 2013-14. (See Characteristics of ongoing studies).
A total of eight papers therefore met the eligibility criteria for inclusion in the current 2012 update of the review (Arick 2005; Blanshard 1993; Brooker 1992; De Nobili 2008; Ercan 2005; Fraser 1977; Lesinskas 2003; Stangerup 1992).
See Characteristics of included studies table.
Two trials included only participants with bilateral OME confirmed by tympanometry (Blanshard 1993; Fraser 1977). Six trials (Arick 2005; Brooker 1992; De Nobili 2008; Ercan 2005; Lesinskas 2003; Stangerup 1992) included participants with both unilateral and bilateral OME. Diagnosis of OME was confirmed by tympanometry alone in two studies (De Nobili 2008; Stangerup 1992), tympanometry and pure-tone audiometry in one study (Lesinskas 2003), tympanometry and otoscopy in one study (Ercan 2005), tympanometry, otoscopy and audiometry in one study (Brooker 1992) and audiometry and otoscopy in one study (Arick 2005). Stangerup and Blanshard only included participants with a diagnosis of bilateral OME of at least three months duration. The diagnosis of OME by tympanogram varied from study to study, with some defining types B and C2 as OME (Blanshard 1993; Stangerup 1992) while others only included type B (Brooker 1992; Ercan 2005).
Seven of the eight trials studied children aged between three and 16 years (Arick 2005; Blanshard 1993; Brooker 1992; De Nobili 2008; Ercan 2005; Fraser 1977; Stangerup 1992). One trial included participants aged between 16 and 75 years (Lesinskas 2003).
The included studies used a range of interventions and outcome measures. Of the eight studies included, three (Blanshard 1993; Ercan 2005; Stangerup 1992) used a classic Otovent®, two (Brooker 1992; Fraser 1977) used a carnival blower + balloon and three (Arick 2005; De Nobili 2008; Lesinskas 2003) used Politzer devices. Seven out of the eight studies used observation without treatment as their control; the exception was Ercan who maintained nasal saline irrigation in the control. The duration of the intervention was variable, with a minimum of 10 days (Lesinskas 2003) to a maximum of seven weeks (Arick 2005). The number of times per day the autoinflation procedure was designed to be carried out was relatively consistent: three times per day (Blanshard 1993; Brooker 1992; Ercan 2005; Stangerup 1992) or two times per day (Arick 2005; Lesinskas 2003). Duration and intensity of intervention specific to each trial can be obtained from the Characteristics of included studies table. One study (Fraser 1977) used a three-factorial design to compare autoinflation, ephedrine drops and Dimotapp
Outcomes and follow-up
Tympanometry was the most commonly used outcome and was reported in four trials, as a dichotomous variable (improvement yes/no) (Blanshard 1993; De Nobili 2008; Stangerup 1992) and as a continuous variable (difference in mean middle ear compliance) (Fraser 1977). Three trials (Brooker 1992; Ercan 2005; Lesinskas 2003) reported dichotomous, composite measures obtained from tympanometry combined with other outcomes. Lesinskas reported a total score that combined pneumotoscopic appearance, tympanometry, patient complaint and audiometry. Brooker reported improvement in both tympanometry and audiometry as a single, combined outcome measure. Audiometry was reported as a separate outcome in three trials. One of these (Blanshard 1993) reported a dichotomous change in pure-tone audiometry threshold at frequencies between 250 Hz and 4000 Hz, while the other two trials reported it as a continuous outcome (mean change in pure-tone audiometry thresholds at frequencies between 500 Hz to 4000 Hz (Arick 2005) or 500 Hz to 2000 Hz (Fraser 1977)). Other outcomes reported were tympanic membrane mobility on otoscopy (Arick 2005) and clearance of fluid from the middle ear (Blanshard 1993). Six studies recorded outcomes at the end of treatment (Blanshard 1993; Brooker 1992; De Nobili 2008; Fraser 1977; Lesinskas 2003; Stangerup 1992). One trial (Arick 2005) recorded outcomes four weeks after the end of seven weeks of treatment. Three trials (Ercan 2005; Lesinskas 2003; Stangerup 1992) reported longer follow-up times after treatment had finished.
See Characteristics of excluded studies table.
A total of 30 studies were excluded for the following reasons:
- one study did not assess hearing loss by either tympanometry and/or audiometry (Chan 1989); and
- one study considered two different autoinflation techniques, but not against a control (Hanner 1997).
Risk of bias in included studies
We assessed studies for risk of bias in four areas: random sequence generation, allocation concealment, blinding of outcome assessment, incomplete outcome data and selective reporting. We also assessed the quality of the intervention in terms of equal provision of care for all groups and treatment adherence. We deemed none of the included studies to be of high quality (see individual entries in the Characteristics of included studies table, Table 1 and Figure 1).
|Figure 1. 'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.|
Only one study fully commented on allocation concealment and we classed this as low risk (Lesinskas 2003). The remaining studies made no comment and so we classified them as unclear risk.
Four studies presented data showing the different arms of the study to be comparable on prognostic characteristics (Arick 2005; De Nobili 2008; Fraser 1977; Stangerup 1992). One study described distribution of age and sex as being similar between arms at baseline (Ercan 2005). One study (Blanshard 1993) commented that groups showed significance only in age distribution and exposure to smoking, while two studies (Brooker 1992; Lesinskas 2003) described as randomised controlled trials did not present data on the baseline characteristics. Confounding was not an issue in any trial.
Only one trial (Arick 2005) reported that the participating doctors and the outcome assessors were blinded, although with patients not blinded it is likely the patient's status could have become known to the assessors.
Follow-up was predominantly low risk: four studies had no attrition (Arick 2005; Brooker 1992; De Nobili 2008; Lesinskas 2003) and three further studies had less than 15% attrition (Blanshard 1993; Ercan 2005; Fraser 1977). We assessed only one study (Stangerup 1992) as high risk, with 42% attrition in the experimental arm and 67% attrition in the control arm.
No trials showed any evidence of selective reporting.
Reported adherence to the use of autoinflation was variable; one trial reported 98% compliance (Arick 2005) while another reported 45% of participants as having high compliance, 43% low compliance and 12% being unable to use the autoinflation device (Blanshard 1993).
Effects of interventions
Eight studies met the inclusion criteria for this review and compared autoinflation with control.
Combining data proved difficult due to a lack of consistent reporting of a common single outcome measure. Hence, post hoc, we also analysed a composite measure of any outcome signifying improvement (as defined in the individual studies). We analysed the data in two time frames: up to one month and more than one month.
Improvement in tympanogram (to type C1 or A tympanogram)
Of the five studies that included tympanometry as a discrete outcome at less than one month, three favoured treatment (Blanshard 1993; Ercan 2005; Stangerup 1992) and two favoured control (Brooker 1992; De Nobili 2008). Within these five studies, at less than one month, autoinflation showed a non-significant improvement in tympanometry as defined by type C2 and B tympanograms (relative risk of improvement (RRI) 1.47, 95% confidence interval (CI) 0.69 to 3.13), with substantial heterogeneity (I
A sixth study (Fraser 1977) used mean change in middle ear pressure as the tympanometry outcome and was not included in the pooled estimates. Fraser found no significant difference between control and intervention groups.
Four studies followed participants for longer than a month (Blanshard 1993; De Nobili 2008; Ercan 2005; Stangerup 1992). In these studies autoinflation produced a non-significant change in tympanometry compared to control (RRI 1.22, 95% CI 1.00 to 1.49, I
Differences in hearing level on pure-tone audiogram (average 10 dB improvement over the frequencies 250 Hz to 2000 Hz)
Two studies included an average improvement of 10 dB or more in the threshold for pure-tone audiometry in a range of frequencies from 250 Hz to 2000 Hz as a discrete outcome measure (Blanshard 1993; Brooker 1992). The estimated recovery was worse in the autoinflation group (RRI 0.80, 95% CI 0.22 to 2.88, I
Improvement measured as a composite of change in tympanogram and/or audiometry (post hoc)
We combined seven studies using a dichotomous outcome of 'improvement' in either tympanogram or audiometry (less than one month: Blanshard 1993; Brooker 1992; De Nobili 2008; Ercan 2005; Lesinskas 2003; Stangerup 1992; and over one month: Arick 2005; Blanshard 1993; De Nobili 2008; Ercan 2005; Lesinskas 2003; Stangerup 1992). The relative risk of improvement was 1.89 (95% CI 0.83 to 4.31, I
Subgroup analysis: type of intervention
We carried out a subgroup analysis based on the type of intervention judged by blinded assessment. Five studies (Blanshard 1993; Brooker 1992; De Nobili 2008; Ercan 2005; Stangerup 1992) provided information on type A interventions (classic Otovent® or carnival blower with balloon) and two (Arick 2005; Lesinskas 2003) provided information on type C interventions (Politzer and modified Politzer). Analyses showed a positive, statistically significant effect using a Politzer device under one month (RRI 7.07, 95% CI 3.70 to 13.51) ( Analysis 3.3) and over one month (RRI 2.25, 95% CI 1.67 to 3.04, I
Adverse effects of autoinflation and compliance
None of the studies demonstrated a significant difference in the incidence of side effects between the control or intervention groups. The side effects/complications studied included incidence of middle ear infection, eardrum perforation, attacks of otitis media, upper respiratory tract infections, tonsillitis, use of antibiotics, middle ear effusion, insertion of grommets and enlarged adenoids. One study (Lesinskas 2003) reported that one patient stopped the treatment due to the pain caused by the procedure.
Six of the eight trials reported compliance. Four studies assessed the adherence to the intervention as 'high compliance' (as defined by the study authors) ranging from 45% to 100%. Two studies reported 'good' or 'satisfactory' compliance. One study commented that "some children with OME found doing the Valsalva manoeuvre painful" (Blanshard 1993), but other than Lesinskas 2003 no other study made any comment.
Data on reported side effects and compliance for each individual trial can be found in the table of 'Side effects and compliance' ( Table 1).
Presence or absence of fluid in middle ear cavity
One study reported otoscopy findings (Blanshard 1993) and demonstrated significantly higher clearance of free fluid from the middle ear at one and two, but not at three months, in the intervention group compared to the control group (P < 0.05).
Lesinskas 2003 presented outcomes as a pooled score of pneumotoscopy, tympanometry, patient complaint and audiometry. At day 10 and day 60 a significant overall improvement was demonstrated in the autoinflation group compared to the control group (P < 0.001). We were unable to obtain data on individual outcomes. Blanshard 1993 analysed low compliance and high compliance subgroups within the intervention. Arick 2005 reported improvement in tympanometric peak pressure (TPP) and hearing recovery (used as primary outcome of improvement in the pooled estimate).
None of the studies reported developmental test results, behavioural test results, language and speech development, quality of life or family function.
Overall autoinflation appears to have a beneficial effect on the resolution of glue ear, with a low incidence of adverse side effects ( Table 1). Six of the eight studies showed beneficial effects at least in the short term (maximum duration of follow-up in the included studies was three months), while one of the two that reported a negative effect in the short term reported a positive effect in the long term (two months). This is in keeping with a previous systematic review on the use of autoinflation (Reidpath 1999). This review concludes that the evidence for the use of autoinflation in the treatment of glue ear in children is conflicting but suggests that it may be of clinical benefit in the short term.
There are several limitations in the collection and interpretation of the findings. The type of device used may be critical in determining the benefits of autoinflation. Pooled effect estimates suggest Politzerisation to be more effective than the classic Otovent® device or carnival blower with a balloon attached. However, both papers that used Politzerisation as the method of autoinflation had slightly different entry criteria to the rest. They were deemed suitable for inclusion in this review as they both used multiple criteria to make the diagnosis of otitis media with effusion (OME) (hearing loss, tympanometry and, in one paper, otoscopy). Information in this area is sparse. A study (Hanner 1997, not part of this review) comparing two different methods of autoinflation (Valsalva versus Otovent®) did not identify a difference between them, however a reanalysis of their data showed a significant difference in favour of Otovent® (relative risk of improvement (RRI) 1.70, 95% confidence interval (CI) 1.08 to 2.69).
Duration of treatment and follow-up also appear to be factors in determining the effects of autoinflation. Autoinflation seems more beneficial in the short term (one month or less of treatment) than in the longer term. Paradoxically the meta-analysis of composite outcomes shows the relative risk of improvement at one month or less to be non-significant (RRI 1.89, 95% CI 0.83 to 4.31) whilst the relative risk of improvement at more than one month is significant (RRI 1.74, 95% CI 1.22 to 2.50). The results are affected by the two studies that found autoinflation to be markedly worse than control at least in the short term (Brooker 1992; De Nobili 2008), leading to significant heterogeneity (I
The combined effects show a mixed pattern of autoinflation, with both statistical and clinical heterogeneity. Considerable heterogeneity may be due to the populations, the intervention or the outcome definitions. One study (Brooker 1992) found autoinflation to be worse than control; another (De Nobili 2008) showed it to be worse in the short term (12 days) but better in the long term (two months). The children in Brooker's study are comparable to those in other studies in terms of age, but there are few data with which to compare the prognostic characteristics of the control and intervention groups in this study with the other studies. For De Nobili 2008 it is difficult to assess if the assumption made about the change in tympanogram for ears is potentially affecting the results (no data on improvement at ear or individual level are provided).
Brooker's populations may be different to those of other studies in that the proportion of controls recovering in this trial was much higher than in any other studies. A possible reason for this is the outcome reported, which is a composite of improvement in pure-tone audiogram and tympanogram. Another difference is that it is suggested in the paper that only children with type B tympanograms were included, but this was not clearly stated. Hence, heterogeneity could be explained if autoinflation is helpful to children with type C2 but not type B tympanograms. Subgroup analyses of the Blanshard 1993 and Stangerup 1992 papers suggest that this might be the case (although the effect size is not statistically significant in either subgroup). Finally, Brooker did not use a standardised Otovent® device, instead using a carnival blower and balloon. This may be a less effective or even harmful method of autoinflation, however the independent rater classified this intervention to be equivalent to Otovent®.
The studies mainly gave information on tympanometry and composite outcomes. Autoinflation does not appear to improve pure-tone audiometry thresholds. Only one study reported otoscopy findings so it is difficult to comment on this outcome. Reduction of hearing loss is an important outcome for children and parents, but there was little evidence within the included studies to address the impact of hearing loss. The development of a quality of life questionnaire specifically for children with hearing loss means that the impact of hearing loss could be assessed in future studies (Umansky 2011).
However, it appeared that autoinflation was beneficial at one and two months. A large primary care study is in progress (Williamson 2011) which will report on one and three-month outcomes.
Adverse events do not appear to differ significantly between intervention and control groups. Data for compliance were variable. There is some anecdotal evidence that some children, possibly in selected groups, do get pain when undertaking autoinflation and that continued enthusiasm for the procedure over time can wane. Blanshard 1993 comments that "significant improvement was seen only in those who used the treatment with a high compliance for the duration of the study". More details on adverse events and compliance are needed from trials, along with evidence on how to achieve good compliance.
The heterogeneity of trial designs, inclusion criteria, interventions, outcome reporting and time scales makes it difficult to draw robust conclusions. Another factor that affects validity is the variation in the level of data analysis. Most studies used ears as their data analysis unit, but some used patients. To overcome this problem we estimated the within-patient correlation. Furthermore, some studies reported their outcomes as continuous variables and some as dichotomous variables. Definitions of improvement varied in the studies that reported dichotomous outcomes. The quality of the trials was variable, in particular none of the studies used blinding of participants or assessors and some studies used inferior methods of randomisation.
One of the major concerns in children with glue ear is possible developmental sequelae; none of the included papers looked at this issue. Currently there are no proven treatments that affect speech and language development, including grommets (Browning 2010). Therefore, it is not possible to say whether the statistically significant improvements in tympanometry translate into clinically significant benefits. This is an important consideration when reviewing treatments for a self limiting condition such as otitis media with effusion; there is a real possibility that watchful waiting without intervention may have similar long-term outcomes to other interventions.
Implications for practice
Although this analysis does not enable us to make any firm conclusions, the evidence appears favourable at least in the short term. Therefore, it may be reasonable to consider autoinflation whilst awaiting natural resolution of otitis media with effusion (OME). One particular concern is the level of adherence: are the child and parent willing to tolerate the inconvenience of the procedure? A recent Cochrane review has concluded that the benefits of grommets for OME compared to watchful waiting are small and then diminish over time (Browning 2010). Currently most autoinflation interventions are instigated in secondary care. With increased awareness and the ability to instruct and educate parents and children in their use, this intervention would be well suited to primary care and there is ongoing research in this area (Williamson 2011). The use of a Politzer device appears promising, however due to the limited amount of information available it is difficult to make any definitive recommendation. None of these trials provided data on long-term sequelae such as language development. However, the trials of ventilation tubes suggest there is little long-term impact achieved through treatment of otitis media with effusion (Browning 2010).
Implications for research
All of the studies we included were small and of limited duration of treatment and follow-up. Further studies with larger numbers of participants and longer duration of intervention, completed in primary care, are much needed. Studies need to provided further evidence on adverse events and treatment compliance. Further studies should also consider the impact of autoinflation on both developmental and hearing loss outcomes in children with otitis media with effusion. The heterogeneity suggests that the method of autoinflation, including the device used, may be important and this merits further investigation. Studies looking at factors predicting which children may benefit from autoinflation would help in identifying subgroups of children for whom autoinflation could be recommended. Future work could also include the possibility of investigating whether autoinflation is effective in preventing disease relapses in children prone to OME.
We are grateful to all authors who supplied us with additional details regarding their studies. We would like to thank Carolyn Dorée for creating the search strategy, running the literature search and extracting the relevant articles. Also, thanks to Martin Burton who provided expert advice throughout the review, particularly in the classification of interventions studied in this review. We would like to thank Daniel Reidpath and Chris Del Mar for allowing us to update and extend their systematic review. We are particularly grateful to Jonathan Blanshard for allowing us access to his original data. Thank you to Gemma Sandberg for running the literature searches for the 2013 update and to Jenny Bellorini for all her advice and support. Thank you also to Jayne Haynes, co-author of the original (2006) version of the review.
Data and analyses
- Top of page
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Differences between protocol and review
- Index terms
Appendix 1. Otovent® and a home Politzerization device (EarPopper™)
Both interventions described here involve inflating the middle ear by blowing air up the nose during the act of swallowing. The main difference is that one intervention is active with the individual (child or adult) inflating a balloon (Otovent®) and then allowing it to deflate; while the other one is passive with air being delivered as a constant flow into the nose (Politzer device). The patient swallows while air is flowing in. During the swallow the air is diverted up the Eustachian tube ventilating the middle ear. There are (at least) two commercial products available, Otovent® (Figure 2) and a handheld, battery-operated device called the EarPopper™ (Figure 3) that allows Politzerisation to be performed at home.
|Figure 2. Child demonstrating the use of the Otovent® device|
|Figure 3. Young woman demonstrating use of modified Politzer device (EarPopper™)|
Appendix 2. Search strategies
Last assessed as up-to-date: 12 April 2013.
Review first published: Issue 4, 2006
Contributions of authors
Rafael Perera: trial selection, data extraction, statistical analysis and writing review.
Jayne Haynes: design of data extraction sheet, trial selection, data extraction and writing review.
Paul Glasziou: basic protocol, compilation of trials, trial selection and writing review.
Carl Heneghan: data extraction and writing review.
Rafael Perera: data extraction, statistical analysis and updating review.
Julie McLellan: trial selection, data extraction and updating review.
Ian Williamson: trial selection, data extraction and updating review.
Declarations of interest
Sources of support
- None, Not specified.
- None, Not specified.
Differences between protocol and review
There has been one main change to the protocol since the original review. We have adopted the Cochrane 'Risk of bias' assessment tool which offers a clearer assessment of study bias across several domains and allows the reader to view how judgements were arrived at.
Medical Subject Headings (MeSH)
*Eustachian Tube; Acoustic Impedance Tests [methods]; Air; Hearing Loss [etiology; *therapy]; Insufflation [instrumentation; *methods]; Otitis Media with Effusion [complications; *therapy]; Pressure; Randomized Controlled Trials as Topic; Valsalva Maneuver
MeSH check words
Adult; Child; Humans
* Indicates the major publication for the study